The GEM® PremierTM 3000 with Intelligent QualityManagement (iQMTM):Features, Technical Description & Statistical Validation

Kevin Fallon, Ph.D.Sohrab Mansouri, Ph.D.

9CRITICALCARE

The GEM® PremierTM 3000 with Intelligent QualityManagement (iQMTM):Features, Technical Description & Statistical Validation

Kevin Fallon, Ph.D.Sohrab Mansouri, Ph.D.

3Table of Contents

I. Features Page 5

II. Technical Description Page 10

III. Statistical Validation Page 17

IV. Glossary of Acronyms Page 21

Appendix Page 22

5I. Features

The new Intelligent Quality Management system (iQMTM) from InstrumentationLaboratory (IL), is a fully automated Quality Assurance system for use on IL’sGEM Premier 3000 analyzer. iQM replaces the need for conventional external Quality Control (QC) with its combination of software, Process Control (PC) Solutions, and Calibration Validaton Product (CVP) components.

Designed to ensure at least the same level of quality as traditional QC methods,iQM significantly reduces time and costs of performing QC, while helping to assure regulatory compliance and improving the quality of test results.

iQM monitors performance of all critical components of the GEM Premier 3000.The "intelligent" quality management system:• Validates cartridge integrity• Identifies changes that affect analytical performance and potential

failure patterns• Automatically performs corrective actions• Automatically documents the failure and corrective actions taken

Unlike external forms of QC, iQM monitors performance real-time, conducting:• System checks• Sensor checks• Stability checks• Pattern checks

System checks include fluidics checks to test sample integrity, reagent solutions, and the peristaltic pump. Additionally, system checks test mechanical components of the cartridge, such as the valve position and sample arm position, as well as the instrument heater block and electronics.All system checks are performed on an ongoing basis.

What It Is

What It Does

How It Works

6

If any system failure is found, iQM confirms the nature of the failure, then either rejects the cartridge or halts instrument operation.

Sensor checks are performed using three different solutions, each at differentintervals. PC Solution B is measured after every patient sample, and in the event the instrument is unused, every half hour. Sensor outputs are monitored in PC Solution B every thirty seconds. During these checks, the drift is analyzed to assess the difference between a measured value and the expected value.

To further test for drift and slope, iQM uses fresh PC Solution A every four hours, and fresh PC Solution C every twenty-four hours. iQM confirms any sensor failure and either suppresses the result of a failed parameter, or disables any sensor that shows persistent failure.

Process stability checks are also performed in order to verify the process control solution stability during the life of a cartridge. In the event of processstability failure, the cartridge is rejected. (Solutions tend to be very stable; the feature simply further ensures accuracy and precision in test results.)

Last, iQM performs Failure Pattern Recognition (FPR) Checks. Common sources of failure, such as micro-clots on sensors, any sensor malfunction, or sensor interference, leave "fingerprints" on the sensors. These fingerprints are read as patterns that iQM can detect and attempt to correct. For example,in the event of a micro-clot, iQM will automatically perform a special rinse. If it cannot correct the problem, the sensor will be disabled. Additionally, iQM can use FPR to warn of the presence of interference in a sample. See "Technical Explanation" and "Statistical Validation" for detail and supportingdata.

Designed specifically to assure accurate and precise test results, iQM exemplifies the simplicity and ease-of-use features associated with the GEM Premier 3000 system. The following unique elements make iQM possible on the GEM Premier 3000:• All analytical components (e.g., sensors, solutions, tubing, sampler)

of the GEM Premier 3000 system are included in a self-contained, multi-use, disposable cartridge (PAK).

Why It Works: Compatibility with the GEM Premier 3000

7

• The GEM PAK is a completely closed system. The user cannot introduce changes to the analytical system during the three-week use-life of the PAK.

• The GEM PAK constitutes a "run," per the National Committee for Clinical Laboratory Standards (NCCLS) definition (the period of time during which an analytical system is expected to be stable).

• The GEM PAK solution values are traceable to NIST standards and stated in each PAK barcode.

• Each sensor and each lot of solutions manufactured by IL are functionally tested prior to PAK assembly. In addition, a representative sample of each PAK batch is functionally tested. Therefore, prior to release of each batch of PAKs, IL is able to test the complete analytical system, includingthe actual combination of analytical elements (e.g., sensors, solutions, tubing, sampler) that will eventually be used by the operator.

• The GEM Premier 3000 analyzer records comprehensive information about all used PAKs, including all calibrations and millivolt (mV) sensor readings. This has allowed IL to study and understand the patterns generated by any PAK failure.

• Immediately after insertion of the iQM PAK into the GEM Premier 3000, and before the system will report any results, an external CVP is runto validate the integrity of the PC Solutions and overall performance of the analytical system.

In order to assure quality results from a blood gas analyzer, hospital staff routinely perform daily liquid QC. This process, as well as monitoring, interpreting and documenting the results, places a significant burden on hospital staff. Traditional QC methods pose many challenges, which may even compromise the quality and accuracy of test results.

For example, with traditional QC:• Manual process requires skilled labor• Aqueous QC ampoules must be run immediately due to the volatility

of O2 and CO2• Corrective action involves potentially complex processes and skills• Manual documentation is needed for regulatory compliance• Discrete process may leave errors undetected until a QC or a series

of QCs are run (after hours or days)

The alternative QC methods advocated by various manufacturers of criticalcare analyzers, namely on-board automated quality control or electronic QC (EQC), are still inadequate in addressing all of the issues previously mentioned.

On-board automated quality control systems are not capable of continuallymonitoring the system functionality and do not address the manual correctiveaction and documentation issues. EQC only confirms the operation of the instrument’s measurement circuitry and its computational components. However, EQC fails to provide information about the chemical measurementprocess, which includes the sensors and reagents, and it is a manual, discrete process.

iQM has been developed to address all of these issues. iQM is an active quality process control program designed to help ensure that the GEM Premier 3000 analyzer provides reliable results at all times. The quality control process is an integral part of system operation, and reduces time to error detection from hours to minutes.

Why It's Better Than Conventional QC Methods

8

iQM continuously monitors operation of the entire testing process including,sensors, fluidics and electronics, and automatically performs and documentscorrective actions upon detection of an error. Running traditional liquid quality controls are no longer needed.

iQM cartridges utilize PC Solutions A, B and C to perform multiple functions: • To indicate if the system has drifted from the reference points established

after cartridge insertion and verification with CVP• As a reference, to eliminate natural sensor drift (“calibrate”)• To identify abnormal sensor behavior drift and trigger FPR software

to determine the cause of the failure, and initiate corrective action• To determine the status of the analytical system, rather than a single

analysis of an external quality control sample. This is achieved by analyzing the collective values of the three PC Solutions, tested hundreds of times during the life of a PAK, along with FPR.

The perception that the same solution cannot be used for two different functions most often originates from experience with other types of diagnosticequipment, such as clinical chemistry systems. In the past, calibrators and controls required manual reconstitution. Reagents were also reconstituted and changed, sometimes within the course of a day. Tubing changes and other maintenance functions that may affect system performance, were also performed routinely. A separate control material was a check that the calibrator was reconstituted correctly, or had not expired, and that other changes to the system did not alter calibration. The GEM Premier 3000, as previously discussed, is a closed system. The operator cannot change anyof the internal cartridge components, and the cartridge will not be acceptedby the system if the expiration date has past. Therefore, prior concerns thatresulted in separate solutions being used for calibrators and controls do notexist with the GEM Premier 3000 iQM cartridge.

PC Solutions play important roles within the iQM cartridge. Therefore, a Process Stability check is performed on the solutions, in the very unlikely event that all PC Solutions deteriorated at the same rate. The Process Stability check is a method to verify PC Solution stability throughout thecartridge use-life. The measured oxygen in PC Solution A during the use-lifeis compared to the very first measured PC Solution A oxygen value during warm-up. The delta must be within allowable limits. The PO2 in PC Solution A is used for the Process Stability check because:

Calibrators and Controls

iQM Traditional QC Electronic QC Auto QCContinuous control Intermittent control Intermittent control Intermittent controlAutomatic corrective Manual corrective Manual corrective Manual corrective action action action actionAutomatic Manual Manual Manual documentation documentation documentation documentation Automated Labor-intensive Manual operation Manual maintenance operation operation of external modulesIntegrated in GEM Additional QC Additional QC Additional QC and/orPremier 3000 PAKs materials required simulators required hardware requiredTotal Analytical May leave errors Electronic component May not check the System QC undetected check only full fluidic pathway

9

• Oxygen is considered the most sensitive parameter for detecting deterioration in PC Solutions, since they have no oxygen-buffering capacity.

• The process of measuring oxygen in PC Solution A utilizes all three PC Solutions. Therefore, deterioration in any of the PC Solutions will affect the measured oxygen in the PC Solution A.

In the event of Process Stability failure, the cartridge will be rejected.

There are PC Solutions used to perform additional sensor checks that do not perform a secondary function for a particular sensor within the cartridge.However, only pH, PCO2 and PO2 sensors require the additional pattern check for identifying malfunctions that are very rare and very slow in progression. Therefore, a PC Solution not used in the initial cartridge verification process is used to further evaluate the performance of these sensors. PC Solution C serves this purpose for pH and PCO2, and PC Solution A provides an additional measurement for PO2.

The GEM Premier 3000 with iQM initiates a new paradigm in quality management. It is the first clinical instrument to apply the “Process Control”approach in the laboratory setting: determine the quality requirements, identify the potential sources of error, and design a method of control for each potential failure.

Statistical analysis of iQM compared to traditional QC has shown that iQMis better than traditional controls at detecting failures that could result in clinically significant errors. iQM can detect more errors, and detect themsooner.

iQM provides laboratories and point-of-care testing locations with the following key benefits:

Quality Assurance• Active, continuous, real-time quality process• Detects and attempts to correct errors immediately• Detects errors sooner than external QC, reducing time to error detection

from hours to minutes• Ensures that the optimal quality control protocol is followed at all times,

regardless of the time of day or operator training

Patient Safety• Helps assure the quality of each patient result• Helps assure that the right medical decision is made for the patient• Prevents reporting sample results when there is a risk of reporting an

incorrect result

Cost Reduction• Eliminates all manual processes associated with traditional quality control• Performs automatically, with no human intervention• Documents failures and corrective actions automatically, with no human

intervention• Eliminates the labor required to maintain a quality control protocol, train

operators, run routine external quality control material (or the maintenanceand refilling of mechanical ampoule beakers), document quality control results, troubleshoot when a failure is identified, and document the failureand corrective action taken

• Produces reports required by regulatory agencies• Reduces inventory management and procurement procedures

iQM Benefits

10

This section provides a detailed description of the GEM Premier 3000 operation, a prerequisite in understanding how iQM functions.

The GEM Premier 3000 system includes two components: the instrument and a disposable cartridge. The cartridge measures pH, PCO2, PO2, Na+, K+, Ca++, Glucose, Lactate and Hematocrit.

The following provides an overview of the cartridge:• All required components for sample analysis are included in the cartridge,

including sensors, sampler, pump tubing, distribution valve, waste bag and PC Solutions. The cartridge components and fluidic path are schematically illustrated in Figure 1.

• The sensor card contains all of the sensors in a gas-tight chamber.• The sensors are monitored with three PC Solutions: A, B and C. These

solutions are pre-tonometered and contain known quantities of the analytestested using known reference standards. The solutions are sealed in gasimpermeable bags with no headspace, allowing their use over a wide range of ambient temperatures and atmospheric pressures. PC Solution B is also used for rinsing.

• There is a fourth bag called "Reference Electrode Solution" that contains silver ion. This solution is pumped into the reference channel in the sensor card to form the Ag/Ag+ reference electrode.

• The sensor card resides in a thermal block, which maintains the temperature at 37°C and provides an electrical interface to the sensors.

• The peristaltic pump moves various fluids (sample, PC Solutions and reference solutions) into the sensor card and eventually to the waste bag.

Understanding GEM Premier 3000 Operation

System description

II. Technical Description

Figure 1. Internal components of the GEM Premier 3000 cartridge

Sampling Position

11

When the cartridge is inserted in the instrument, all information associated with the cartridge is read from the cartridge bar code and recorded. This information includes cartridge type (specifying menu, sample capacity anduse-life), expiration date and PC Solution values. Once the cartridge bar code is validated (new cartridge with valid expiration date), the sensors arehydrated and calibrated within 30 minutes before the analyzer becomes ready for use.

Cartridge quality assurance is achieved primarily by two methods:1. Testing critical components of the cartridge before assembly2. Downloading all cartridge data for investigation

The critical components of the cartridge, sensor card and PC Solutions, aredesigned for testing before cartridge assembly. Unlike sensors in othercartridge-based systems, the GEM Premier 3000 sensor card can toleratemany cycles of hydration and drying without any deterioration in sensor performance. Furthermore, the fluid opening valve in the reagent bag is configured for sampling without affecting its solution composition.

All manufactured sensor cards are tested in an elaborate and automated computer-controlled testing system. The test utilizes PC Solutions similar to those in the cartridge, plus an additional test solution for performanceevaluation. Only the sensor cards that pass stringent criteria based on sensitivity (slope), accuracy, and drift are assembled in cartridges.

A statistically significant number of reagent bags from a production lot are tested for quality assurance and for assigning analyte values. The value assignment process is traceable to National Institute of Standards and Technology (NIST) standards. The assigned concentration values are included in the cartridge bar-code.

All system and cartridge operations are recorded on the instrument hard drive. The recorded data is very comprehensive. For example, every sensormV, every sample or PC Solution measurement, and every check or alarm isrecorded. This information can be downloaded onto a floppy disk or directlye-mailed from the instrument through the GEMweb™ feature. In case of cartridge malfunction, the user can send the cartridge data to the complaint investigation department at IL for analysis and failure determination.

iQM operation is achieved by adding additional layers of checks to the GEM Premier 3000 operation. iQM uses FPR to identify cartridge malfunctionsthat are currently undetected by internal system and sensor checks.

iQM operation can be summarized as follows:• Monitors performance of the system in real-time• Identifies potential failure patterns• Automatically performs corrective actions• Automatically documents the failure and action taken

Cartridge Operation

Cartridge Quality Assurance

Testing Cartridge Components

Recording Cartridge Data

Understanding iQM Operation

12

iQM operation begins following cartridge verification after warm-up. External CVP is used to independently verify cartridge calibration. Four ampoules of CVP (two levels for pH, blood gases, electrolytes and metabolites, and two levels for hematocrit) must pass before the cartridgecan be used for sample reporting. If one or more of the analytes fails CVP testing, those channels will not be available for sample analysis.

Once cartridge calibration is verified, iQM monitors the status of the calibra-tion during cartridge use-life. Upon detecting a problem, the analyzer auto-matically performs corrective actions.

The following are examples of corrective actions:• Automatically performing a special rinse cycle, then subsequently verifing

cartridge function if micro-clots are detected• Permanently disabling a failed sensor if its functionality cannot be recovered• Rejecting a cartridge for process stability failure• Alerting the operator if an interfering substance in the sample is detected

During cartridge operation, the instrument automatically and continuously performs various checks that can be categorized into four groups: 1. System checks2. Sensor checks 3. FPR checks4. Process Stability checks

System checks include basic functionality of the instrument and the cartridgeExamples of these checks include:• Cartridge fluidic checks to assess sample integrity, detect the presence

of PC Solutions, and verify peristaltic pump functionality• Cartridge mechanical checks, such as proper operation of the distribution

valve and sampler arm• Instrument heater-block checks, such as heater-block temperature and

mV-output threshold from sensors (e.g., outputs at rail)• Instrument electronic checks, such as analog/digital electronic calibration

verification and communication between processors

Any failure in the system checks will lead to a corrective action. The correctiveaction will include verification of the failure followed by one of the followingsteps:• Cartridge rejected, in case of cartridge-related failure• Instrument operation halted, in case of instrument-related system failure

Sensor checks address functionality of the sensors. PC Solutions A, B and C are automatically brought into the sensor card at various intervals to verify sensor operation. The solution that is residing in the sensor card is measuredand the drift is determined. (Drift is the delta between the measured and expected value.) The expected values are obtained from either the cartridgebar-code or from prior PC Solution measurements.

Sensor checks are performed with the following frequencies:

• PC Solution B is measured most frequently. Fresh B is measured, at minimum, every 30 minutes or after every sample. Furthermore, PC solution B is measured every 30 seconds while residing in the sensor card in between fresh B measurements.

System Checks

Sensor Checks

13

• PC Solution A is measured, at minimum, every 4 hours. All sensor slopevalues are also measured and checked. Slope, which is an indicator of sensor sensitivity, must be within allowable limits.

• PC Solution C is measured, at minimum, every 24 hours. PC Solution C isprimarily used for measuring low-level oxygen and for verifying performanceof the pH and PCO2 sensors. It is also used for conditioning the glucose and lactate inner membranes, which provide rejection of interferencecompounds present in blood.

It should be noted that the PC Solution B and PC Solution A measurementsare transparent to the operator. The measurement can be interrupted at anytime to run a sample (except the A measurement within the first 4 hours of cartridge life, which is uninterruptible). The PC Solution C measurement is uninterruptible, but the operator has the option of selecting the exact time of day for performing PC Solution C.

The process of verifying sensor operation by measuring PC Solutions is very similar to the process of sample measurement. As indicated in Figure 2, the sample path and the PC Solution path into the sensor card is identical. In addition, the algorithms for calculating the PC Solution values are similar to those for sample calculations using prior calibration values. After performing sensor checks, the analyzer conducts a number of actions. If all measured values are within allowable limits, the sensors will be calibrated and, as a result, the drifts will be set to zero.

If any measurement or slope value is outside the allowable limits, the following corrective actions will take place: • Parameter result in subsequent sample report will be suppressed• If the failure fits with a known pattern, specific corrective actions will

follow (see FPR checks). If not, fresh solution will be brought into the sensor card for up to two more times and measured

• If the failure persists in two consecutive C measurements, in three consecutive A measurements, or 15 consecutive B measurements, the failed parameter will be permanently disabled

Figure 2. Similarities between sample path and PC Solution path

Home Position

14

FPR checks were developed through years of investigating field complaints. As indicted in the previous section, GEM analyzers (GEM Premier 3000 as well as the GEM Premier) provide comprehensive information about used cartridges. This information, plus in-house investigations about the cause of failures, provided the background for the FPR checks.

Two distinct failure patterns were identified: micro-clot-related failures and certain sensor malfunctions that are not well detected by other internal checks or by external QC. Furthermore, certain interference patterns were also specified.

Micro-clots are small blood clots or fibrin strands that adhere to a sensor and induce a change in sensor characteristics, such as sluggish response or sensitivity change. Micro-clot patterns are distinct for various sensors. Sensor- check failures (drift errors) are used to identify the presence of micro-clots.

In the event of micro-clot pattern detection, iQM automatically initiates a specialrinse cycle, using PC Solution C (PC Solution B is used for normal rinse). PC Solution C has added surfactants for better clot removal. Upon completion of the rinse cycle, iQM checks for a clot pattern on the affected sensor. If a clot pattern still remains, the affected sensor is disabled. If a clot pattern is not detected, the sensor status becomes green (ready for measurement).

Figure 3 illustrates an example of micro-clot formation on the oxygen sensor and the pursuing corrective actions. In this example, the detected pattern

FPR Checks

Micro-Clot Patterns

60

50

40

30

20

10

048.5 49.5

B B B B B

BB

B B B B B B

A

Sample # 18

Sample # 17

Sample # 19 Sample # 20

Sample # 21

A

Clot Verified

Cartridge Use Life (hr)

PO

O

utp

ut (

mV

)2

Clot DetectedClot Removal Verified

Rinse Ready

50.5

B drift

A drift

A & B Sequence

PO

(m

mH

g)

2

100100 110 120 130 140

120

140

160

180

200

Corrected

Measured A

Measured B

Micro-Clot Pattern Detected- Negative B drift error after sample- Positive A drift error

Figure 3. Example of Micro-Clot Pattern for PO2 Sensor

15

was a negative drift error in the PC Solution B oxygen check after sample #19, followed by a positive drift error in the oxygen PC Solution A check which was performed automatically and immediately after the PC Solution Bcheck failure. After confirming the presence of a micro-clot (by the PC Solution A check), the special rinse cycle was initiated. Following the rinse cycle, the removal of the clot was verified by re-running PC Solution B and Achecks. In this example, the entire process of detection, clot removal and verification took about 10 minutes.

Only pH, PCO2 and PO2 sensors need an additional pattern check for identifying malfunctions. Existing sensor checks are found adequate for detecting any malfunction in other sensors. It should be noted that normal sensor malfunctions for pH and PCO2 sensors are identified with the existing sensor checks. The specific malfunctions for which iQM is checking in thesesensors are very rare and very slow in progression. Therefore, the daily PC Solution C check is adequate in detecting these malfunctions.

In case of a sensor malfunction pattern, the affected sensor is permanently disabled by iQM.

Two interference patterns are checked, one for positively charged lipophilic compounds, and one for negatively charged lipophilic compounds. These compounds can cause false readings for a few of the ion-selective electrodes.Benzalkonium Chloride is a good example of a positively charged lipophilic compound that can cause false-elevated readings for sodium and ionized calcium.

Sodium Thiopental is an example of a negatively charged lipophilic compound.For an interference pattern to be positively identified, all conditions must be met. In the event of interference pattern detection in a sample, iQM notifies the user.

Figure 4 illustrates an example of Benzalkonium interference on the ionized calcium sensor. The detected pattern was a positive drift error in the ionized calcium PC Solution B check, along with certain drift movements in the sodium and potassium sensors. Following the interference detection, PC Solution B is performed at a higher than normal frequency (every 3 minutesversus every half-hour).

Sensor Malfunction Patterns

Interference Patterns

3021

Sample

Sample with Benzalkonium

B drift error

Cartridge Life (hours)

Sen

sor

Out

put (

mV

)

21.5 22 23 23.522.5 24

35

40

45

50

Figure 4. Example of Interference Caused by a Positively Charged LipophilicCompound

16

Process Stability check is a method of verifying the PC Solution stability throughout the cartridge use-life. The measured oxygen in PC Solution A during the use-life is compared to the very first measured PC Solution A during warm-up. The delta has to be within allowable limits.

The PO2 in PC Solution A is used for the process stability check because: • Oxygen is considered the most sensitive parameter for detecting

deterioration in PC Solutions, since they have no oxygen-buffering capacity

• The process of measuring oxygen in PC Solution A utilizes all three PCSolutions, so that deterioration in any of the PC Solutions will affect the measured oxygen in PC Solution A.

In the event of process stability failure, the cartridge will be rejected.

Process Stability Check

17III. Statistical Validation

As indicted in the previous sections, drift limits are used as a trigger for sensor and FPR checks, and for subsequent iQM corrective actions. Therefore, drift limits should be optimized for high probability of error detectionand low probability of false rejection. This section explains how statistical control methods are used for evaluation of drift limits.

Drift limits on PC Solutions A and B can be characterized as a single measurement of a control material. Statistical control methods are then used to develop probabilities for error detection and false rejection. This approach allows a comparison of performance expected for iQM with performance of traditional QC procedures.

The method is as follows:• Define the quality requirement in terms of total allowable error (TEa)

- pH, PCO2, Na+, and K+ = CLIA 88 limits- PO2 = 10%- Ca++= 9%- Glucose = 12% or 12 mg/dL- Lactate = 15% or 0.4 mmol/L- Hematocrit = 2%

• Define method performance- Mean and SD values are obtained from the data collected from

24 cartridges representing a wide variety of uses from customersin the field and in-house tests

• Predict QC performance- Calculate Method Sigma = TEa/SD- Calculate Control Limit = Drift Limit/SD- Determine probability of false rejection (Pfr) from normal probability

distribution (from tables of areas under normal curve, or z charts)Pfr = Prob(z ≥ Control Limit)

- Determine probability of error detection (Ped) with 95% confidence from normal probability distribution

Ped = 1 – Prob(z ≥ (Method Sigma – Control Limit – 1.65)) - Calculate Average Run Length for rejectable quality

ARLr = 1/Ped- Determine average detection time (unit of time for detecting error that

can be compared to traditional QC)Average detection time = ARLr x sampling time

As specified previously, sampling time for Solution A is between 1 to 4 hoursand for Solution B is between 3 and 30 minutes (3 minutes when there is a sample in between B measurements). For a given TEa, the drift limits must provide a high probability of error detection (Ped ≈ 1) and a low probabilityof false rejection (Pfr ≈ 0). Tables 1 and 2 provide a summary of the analysisof the drift limits for PC Solutions A and B.

Results of the drift limit analysis indicate that the probability of false rejection is close to zero for all parameters in PC Solutions A and B. PC Solution B is the primary means for error detection because of high measurement frequency.Probability of error detection is high in the PC Solution B. Even for sodium and glucose with low Ped values, the average error detection time is within 10 minutes.

Statistical Evaluation of Drift Limits2

Statistical Method2

18

Table 1. Statistical analysis of drift limits for PC Solution A

pH PCO2 PO2 Na+ K+ Ca++ Glu Lac HctMean 6.90 64 119 102 6.7 2.63 141 3.1 25SD 0.006 1.25 1.65 0.73 0.043 0.034 3.71 0.09 0.05TEa 0.04 5 11.89 4.0 0.50 0.25 17.0 0.5 2Method Sigma 6.4 4.0 7.2 5.5 11.5 7.4 4.6 5.6 39.2Drift Limit 0.03 4 6 3.0 0.30 0.15 14 0.30 1Control Limit 4.8 3.2 3.6 4.1 6.9 4.4 3.8 3.3 19.6Pfr 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000z for Ped -0.06 -0.85 1.92 -0.27 2.96 1.29 -0.84 0.58 17.96Ped 0.476 0.198 0.973 0.393 0.998 0.902 0.200 0.718 1.000ARLr 2.10 5.05 1.03 2.54 1.00 1.11 4.99 1.39 1.00ATDr (4 hr) 8.4 20.2 4.1 10.2 4.0 4.4 20.0 5.6 4.0

Table 2. Statistical analysis of drift limits for PC Solution B

*Compare with a typical QC program where QC is performed only every 8 hours

pH PCO2 PO2 Na+ K+ Ca++ Glu Lac Hctunit mmHg mmHg mmol/L mmol/L mmol/L mg/dL mmol/L %

Mean 7.40 32 176 144 3.6 1.16 0 0.0 11SD 0.002 0.41 1.37 0.90 0.011 0.009 1.4 0.05 0.09TEa 0.04 5 18 4 0.5 0.10 12 0.4 2Method Sigma 20.0 12.3 13.6 4.5 45.9 11.1 8.9 9.7 23.1Drift Limit 0.03 3 10 3 0.3 0.06 10 0.3 1Control Limit 15.0 7.3 7.3 3.3 27.5 6.8 7.3 6.4 11.6Pfr 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000Ped 1.000 1.000 1.000 0.298 1.000 1.000 0.425 0.942 1.000ARLr 1.00 1.00 1.00 3.36 1.00 1.00 2.35 1.06 1.00Average Detection Time (min)* 3 3 3 10 3 3 7 3 3

The effectiveness and efficacy of iQM error detection was investigated usingthe existing field complaint data files for the GEM Premier 3000 with Phase II cartridge configuration. (Phase II cartridge configuration has the capability to include glucose and lactate). It should be noted that the FPR checks wereoriginally developed from field and in-house data using the GEM Premier 3000 and Phase I cartridge configurations.

The investigation focused on all reported field QC failures from US accounts beginning with the release date of Phase II cartridges in June 2001, to the endof March 2002. (Only QC failures were selected because iQM is expected todetect cartridge malfunctions currently identified by running external quality controls.) Approximately 7000 cartridges were shipped to U.S. GEM Premier3000 accounts during this period.

Summary data is outlined below.• A total of 164 QC-related cartridge failures were reported to IL• From 164 reported QC failures, IL obtained 144 data disks• From 144 obtained data disks, IL confirmed 117 real QC failures

(others either had no QC failure or had failures other than QC)• From 117 confirmed QC failures, 79 passed initial QC out of warm-up,

but failed QC later

Evaluation of Cartridge Malfunction Detection

19

Data from the 79 GEM Premier 3000 cartridges that had confirmed QC failures during cartridge use-life was extensively analyzed. The FPR checkswere applied to the data to determine if iQM could detect any malfunction.

The results are summarized below.• 68 of the data files exhibited micro-clot failure patterns.

- PO2: 46- Hematocrit: 16- Ca++: 2- pH: 1- PO2 and Hematocrit: 2- PO2, Na+ and Ca++: 1

• 6 showed a sensor failure pattern.- PCO2: 4- pH: 2

• 3 demonstrated one QC level at the limit of the acceptable range with a subsequent QC failure.- PO2: 2- Glucose: 1

• 2 had no identifiable failure pattern; both QC failures were marginal and occurred at only one level.- Na+: 1- Ca++: 1

This analysis indicates that FPR was able to detect malfunction in 77 out of 79 reported QC failures. It is interesting to note that in most cases, FPR flagged the failure immediately after the sample that caused the malfunction; while the operator typically became aware of the problem only later when the QC was run and failed. In some cases, QC had been run several hours after occurrence of the malfunction. Furthermore, no false-positive flag wasgenerated by FPR in all 79 investigated cases.

No obvious malfunction was detected in the data from the two cartridges that FPR did not flag but had reported QC failure. All system parameters were found within specifications. Blood results for the parameter with repor-ted QC failure were examined and found reasonable. In both cases, the reported QC failures were marginal and at one level only. It was concludedthat the reported QC failures in those two cartridges could not represent a serious cartridge malfunction. The failures could be considered as a false-positive QC failure.

During a limited distribution phase, iQM was implemented in 20 hospital sites.These selected sites represented high volume locations, many with a history of pre-analytical issues tha may affect patient results. The issues included samples containing micro-clots and samples contaminated withBenzalkonium.

Data from 42 cartridges from the various clinical sites, using various cartridge types, were analyzed at IL. Cartridge files with reported iQM failures were of particular interest in the investigation. Two-thirds of the cartridge data collected showed no cartridge issue, while one-third had at least one reported problem during cartridge use-life. The iQM Corrective Action Report was used to identify any cartridge issue. Additionally, the sensor output data files, the sensor calibration data files, and the system logs were examined to confirm the validity of iQM reported errors.

iQM Error Detection During Limited Distribution

20

iQM correctly flagged the presence of micro-clots in 11 cartridges. Five of the micro-clots were successfully rinsed out, while 6 could not be removed, and the analyzer automatically disabled the affected sensors. There were 5 flags for Benzalkonium interference on the sodium and ionized calcium measurements. The sodium and ionized calcium values in those flagged samples were abnormally high. In all cases, iQM flagged the problem withinminutes of the occurrence.

iQM was very effective in identifying cartridge issues promptly. The automaticcorrective action function of iQM was effective in removing about one-half of the detected micro-clots.

iQM is an active, real-time quality process control program that allows for immediate error detection and correction, thus further enhancing quality assurance in the GEM Premier 3000 analyzer.

Summary

21

ARLr: Average Run Length for Rejectable Quality

CAR: Corrective Action Report

CVP: Calibration Validation Product

EQC: Electronic Quality Control

FPR: Failure Pattern Recognition

iQM: Intelligent Quality Management

IL: Instrumentation Laboratory

NCCLS: National Committee for Clinical Laboratory Standards

NIST: National Institute of Standards and Technology

PC: Process Control

Ped: Probability of Error Detection

Pfr: Probability of False Rejection

QC: Quality Control

SD: Standard Deviation

TEa: Total Allowable Error

IV. Glossary of Acronyms

22

AI. InterferencesThe following substances can potentially interfere with sample analysis:• Benzalkonium Chloride and Benzalkonium Heparin: Arterial lines and

sampling devices coated with these substances may interfere with Na+

and Ca++ determinations, causing falsely elevated Na+ and Ca++ readings.Following sample analysis, and analysis of PC Solution B, if Benzalkonium Chloride or Benzalkonium Heparin patterns are detected, the following message will be displayed on the analyzer, and will persist until acknowledgedby the operator:

Sensor Interference Detected for Na and iCa on last sample likely due to Benzalkonium

The GEM Premier 3000 offers the operator the ability to enable flagging of patient results if an interference pattern is detected. In addition, this option, when enabled,delays the reporting of results until PC Solution B is evaluated following sample analysis. If flagging of patient results is enabled, the following message (plus progress bar) will be presented while the post-analysis PC Solution B check is underway:

Checking for presence of interference and micro clots

This message will remain displayed until the analysis of PC Solution B, is complete. If an interfering substance pattern is detected, the affected blood result(s) will be flagged. In addition, the analyzer will beep three times to alert the operator. The following message disappears only after operator acknowledgment:

Sensor Interference Detected for Na and iCa on last sample likely due to Benzalkonium

• Thiopental sodium may interfere with the Na+, K+, PCO2 and Ca++ readings.Thiopental Sodium is also known by other names, including: ThiomebumalSodium, Penthiobarbital Sodium, Thiopentone Sodium, Thionembutatal, Pentothal Sodium, Nesdonal Sodium, Intraval Sodium, Traoanal, and Thiothal Sodium.

Following sample analysis and analysis of PC Solution B, if the associated pattern is detected in PC Solution B, the following message will be displayedon the analyzer, and will persist until operator acknowledgment:

Sensor Interference Detected for xxxxx on last sample (where xxxx is the analyte or analytes affected)

When flagging of patient results for an interference is enabled, the following message (plus progress bar) will be presented while the post analysis PC Solution B check is underway:

Checking for presence of interference and micro clots

This message will remain displayed until the PC Solution B is complete. Atthat point, processing will continue until results are displayed. If Thiopental Sodium is detected, the affected blood result(s) will be flagged. In addition,the analyzer will beep three times to alert the operator. The following message disappears only after operator acknowledgment.

Sensor Interference Detected for xxxxx on last sample (where xxxx is the analyte or analytes affected)

Appendix

Note: the following information is based on data obtained during the development of iQM cartridges; please refer to the complete GEM Premier 3000 Operator’s Manual (Revison 3) for complete specifications on all available GEM Premier 3000 cartridges.

23

AII. Performance Characteristics SummaryGEM CVP PrecisionPrecision data were generated at IL using GEM CVP: 2 levels for pH, bloodgases, electrolytes and metabolites and 2 levels for Hematocrit.

Based on NCCLS guidelines, the verification material levels were run in singlet once a day for 14 days (twice on day 1) for a total of 15 replicates on each of 9 different GEM Premier 3000 instruments (N=135). The table below lists the combined results of the 9 instruments.

Note: SD is used for pH since differences are so small that %CV would be misleading.

GEM CVP Level 1:

Parameter Mean Day-to-Day %CV (or SD) Total %CV (or SD)pH 7.200 0.005 (SD) 0.007 (SD)PCO2 (mmHg) 70.8 1.39 1.63PO2 (mmHg) 54.5 4.97 5.16Na+ (mmol/L) 129.3 0.46 0.55K+ (mmol/L) 2.90 0.25 0.70Ca++ (mmol/L) 1.493 0.95 1.26Glucose (mg/dL) 46.1 2.23 2.99Lactate (mmol/L) 0.93 4.73 4.87

GEM CVP Level 2:

Parameter Mean Day-to-Day %CV (or SD) Total %CV (or SD)pH 7.640 0.002 (SD) 0.003 (SD)PCO2 (mmHg) 29.9 1.78 1.91PO2 (mmHg) 148.2 1.33 1.93Na+ (mmol/L) 158.7 0.44 0.56K+ (mmol/L) 6.46 0.75 0.98Ca++ (mmol/L) 0.486 1.15 2.06Glucose (mg/dL) 192.8 1.67 1.78Lactate (mmol/L) 5.54 1.85 2.19

GEM CVP Level 3:

Parameter Mean Day-to-Day %CV Total %CVHematocrit (%) 23.4 2.14 2.11

GEM CVP Level 4:

Parameter Mean Day-to-Day %CV Total %CVHematocrit (%) 43.8 1.21 1.23

24

y = 1.0802x - 0.581r = 0.9917n = 281

pH Units Reference

pH

Un

its iQ

M G

EM

pH Method Comparison

7.10

7.100

7.20

7.30

7.40

7.50

7.60

7.200 7.300 7.400 7.500 7.600

y = 1.674x - 2.38r = 0.9839n = 282

mmHg Reference

mm

Hg

iQ

M G

EM

PCO Method Comparison

20

20.0

30

40

50

60

70

80

90

100

30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

2

Arterial, venous, heart bypass and liver transplant blood samples were obtained from patients using heparinized syringes. The table and graphs below demonstrate that the GEM Premier 3000 using an iQM cartridge is statistically similar in performance to a reference analyzer.

Analyte N Slope Intercept r Sample RangepH 281 1.0802 -0.581 0.9917 7.129-7.559 PCO2 (mmHg) 282 1.0674 -2.380 0.9839 25.3-87.5PO2 (mmHg) 282 0.9715 6.990 0.9988 26-489Na+ (mmol/L) 271 0.9801 2.926 0.9584 119-148K+ (mmol/L) 271 0.9743 -0.061 0.9871 3.2-7.4Ca++ (mmol/L) 264 0.9620 0.0396 0.9923 0.86-1.56Glucose (mg/dL) 283 1.0111 8.868 0.9860 66-389Lactate (mmol/L) 279 0.9382 0.163 0.9957 0.49-15.07Hct (%) 284 0.9983 -0.800 0.9600 17-56

AIII. Method Comparison

25

y = 0.9715x - 6.99r = 0.9988n = 282

mmHg Reference

mm

Hg

iQ

M G

EM

PO Method Comparison

0

0

100

200

300

400

500

600

100 200 300 400 500 600

2

y = 0.9801x - 2.93r = 0.9584n = 271

mmol/L Reference

mm

ol/L

iQM

GE

M

Na Method Comparison

115

115

120

125

130

135

140

145

150

155

120 125 130 135 140 145 150 15 5

+

y = 0.9743x - 0.06r = 0.9871n = 271

mmol/L Reference

mm

ol/L

iQ

M G

EM

K Method Comparison

2.5

2.5

3.5

4.5

5.5

6.5

7.5

3.5 4.5 5.5 6.5 7.5

+

26

y = 0.962x - 0.0396r = 0.9923n = 264

mmol/L Reference

mm

ol/L

iQ

M G

EM

Ca Method Comparison

0.7

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

0.9 1.1 1.3 1.5 1.7

++

y = 1.0111x - 8.87r = 0.9860n = 283

mg/dL Reference

mg

/dL

iQM

GE

M

Glucose Method Comparison

50

50

100

150

200

250

300

350

400

450

150 250 350 450

y = 0.9382x - 0.1633r = 0.9957n = 279

mmol/L Reference

mm

ol/L

iQM

GE

M

Lactate Method Comparison

0.0

0.00

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

5.00 10.00 15.00

27

y = 0.9983x - 0.7999r = 0.9600n = 284

% Reference

% iQ

M G

EM

Hematocrit Method Comparison

15

15

20

25

30

35

40

45

50

55

60

25 35 45 55

28

Figure 5. iQM Delta Chart: Components.

12

34

56

712

1314

15

10

9

8

11

}

AIV. iQM ReportsiQM Delta ChartsAn iQM Delta Chart is generated for each Process Control Solution and analyte combination. A Delta Chart, depicted in Figure 5, includes:1. Analyzer (GEM Premier 3000)2. Analyte name3. Analyzer serial number and name, if the analyzer is given a unique

name by the facility4. Nominal target value for the analyte5. Report month and year6. PC Solution 7. The number of times the PC Solution was run each day (If the number

of points exceeds 99, 99 will be displayed)8. The maximum delta between the Process Control Solution target value

and the actual measured value for the selected analyte each day, represented by a short horizontal line

9. The mean delta between the Process Control Solution target value and the actual measured values for the selected analyte each day, depicted by a round, bolded dot

10. The maximum delta between the Process Control Solution target value and the actual measured value for the selected analyte each day, represented by a short horizontal line

11. A vertical line, which connects the maximum, mean and minimum delta points

12. The maximum allowable delta for the analyte13. The minimum allowable delta for the analyte14. The date of cartridge insertion (indicated by an arrow), along with the

cartridge lot number

29

Delta points outside the designated limits will not be shown on the Delta Chart, but will be included in the iQM Corrective Action Report. This is because the GEM Premier 3000 requires the use of static graphs (graphs with ranges that do not change).

Please refer to Figure 6 for graphical representations of the following situations:

1. It is possible to have the minimum and maximum delta values coincide with one another and the mean delta result. This happens when:• The same delta value is obtained for an analyte each time a PC

Solution is run.• Only one delta value is obtained for an analyte through the course of

a day, which is expected for PC Solution C.

2. A minimum or maximum delta result (represented by a horizontal line) maycoincide with a daily mean delta result. An example of this occurring is:• K+ is measured in PC Solution B 90 times in one day• 89 x the delta value = 0• 1 x the delta value = +0.01

In this case, the mean and minimum delta values will be the same (0). The horizontal line representing the minimum delta value will intersect the round, bolded dot representing the mean value. Furthermore, since the minimum delta value is 0, the horizontal line representing the minimum deltaresult will coincide with the 0 horizontal line on the graph.

3. If the maximum and/or minimum delta obtained during a day coincides with the maximum or minimum allowable delta limits, then the horizontal line will “blend” with the upper/lower graph line limits.

Figure 6. iQM Delta Chart: Specific Delta Value Representations.

30

Calibration Validation Product ReportsMonthly CVP reports can be printed. A CVP report contains results for all CVP ampoules run within the specified month.

Figure 7. iQM Corrective Action Report-Components.

13

567

8

9 {

42

iQM Corrective Action Report (CAR)The Corrective Action Report contains information for significant events that occur during the cartridge on-board use-life. The CAR, shown in Figure 7 includes:1. Analyzer, analyzer serial number and name2. Month and year3. Date and time event occurred4. Cartridge Lot Number5. Detected failure description6. Operator ID, if entered (if the function included operator interaction)7. Corrective action description8. Corrective action result9. A Cartridge Removal entry that lists the total number of Process Control

solution adjustments made during the cartridge use-life. The adjustments represent the total number of minor drift errors that are corrected by reanalyzing Process Control solution B. Events, such as interfering substance detection, micro-clot detection, and fatal errors, will be listed and described as individual events on the log.

31

References

Biographies of Authors

1. GEM Premier 3000 Operating Software, Volume 2, Cartridge Internal Operations, Software Version 5.2, Revision 2.7.

2. Westgard JO, Fallon KD, Mansouri S. “Validation of iQM Active Process Control Technology” Point of Care, 2003; Vol 2 (1): 1-7.

3. “Thiopental Interference in GEM Blood Gas Systems” by H. Hanford, Technical Note #112-833.

4. “Tietz Fundamental of Clinical Chemistry” by C.A. Burtis and E.R. Ashwood, Forth Edition (1996), page 828.

Kevin Fallon, Ph.D.Dr. Fallon was a member of Instrumentation Laboratory’s scientific staff for 25 years, most recently serving as the Director of Scientific Affairs. Additionally, he served on the faculty of the University of Texas Medical School and was Director of the STAT Labs at Hermann Hospital in Houston.Dr. Fallon continues to offer his expertise to IL on a consulting basis.

Sohrab Mansouri, Ph.D.Dr. Mansouri has over 20 years of experience in the design, development andimplementation of chemical and biochemical sensors for medical applications.He has spent the past 13 years advancing the GEM technology and is responsible for the development of the Premier and Premier 3000 cartridges.

© 2003 Instrumentation Laboratory - Printed in Italy - Speed 2000 - 06/03

Par

t.N

o 98

0861

9